Hail mary pass since it has been brought up from time to time. AI pilots seem to be able to dive any distance and at very steep (near vertical) angles and then pull out without suffering any structural integrity failure to their craft.

THE REQUEST
Is it possible to mod the AI to either detect the stress levels of the wings and limit their dive accordingly OR limit how far each particular type of craft will steep dive (Spads much longer than N11s, for example) before switching to another tactic so that the AI might more closely simulate how pilots would handle their craft historically during air combat?

THE CLARIFICATION
I know this gets tricky because sometimes the AI fakes a "death spiral" and then pulls out several hundred feet lower in an attempt to lose their pursuer and escape the combat. That is a perfectly understandable move by the AI, and in fact I love that. What I am referring to are the ultra steep, extended dives, often at a slightly cocked angle, that would appear to push a machine past its limits and for which players in the same aircraft find well near impossible to replicate because their wings suffer massive damage if not outright catastrophic failure.

THE HOPE
Success in making the AI pilots fly within similar structural integrity limits of players would keep the dogfights higher in altitude and lasting longer, instead of losing a lot of altitude quickly and fighting down on the deck. It should lead to "more time dogfighting and less time diving. It may well also reduce the number of dogfights that make it to the deck level, since they are more likely to be resolved at higher altitudes."

THE BONUS
If Novice AI were most likely to push their aircraft past its limits and rip their wings off, Average pilots less so and Aces virtually never. Probably impossible, but would be cool if it could be done.

In principle, I am all in favor of this idea, the problem, of course, is that we don't have access to the code that would probably be required to implement your suggestions. Unfortunately, both the FM and the AI flight manuever schemes are encrypted files in WOFF.

Well poo. LOL

But perfectly understandable given all the time and effort the devs put into creating them. Ah well, maybe someday. I think it would take the air combat to yet another higher level of "realism." Then again, maybe the devs had their reasons for not doing this.

Interesting idea, but how do we "communicate" this information to the AI? Short of changing the code, how would the sim and AI "know" that they are in a steep dive that might potentially rip off their wing? I'm sure there is some way for the code to communicate this to the AI, but how do we lowly modders gain access to it?

Another problem is calculating the physics to distinguish between a relatively peaceful and non life-threatening shallow dive vs. a death defying screaming power dive. Would we want the AI to prematurely terminate a relatively shallow dive simply because of the expiration of an artificial time limit?

I wish I could be more helpful, but I am afraid my tiny brain may not be up to the task. Perhaps other modders have some better ideas?

Ha!


Perhaps after WOTR Gold has been released, OBD will unencrypt the AI routines or release modding tools that will enable us to finally fiddle around with this stuff! Can you wait for another five or ten years?

I look at it this way. As a junior pilot you would be surely rapped over the knuckles for "Following an enemy down" by your superiors for several reasons.

1) It's a form of target fixation that leaves you vulnerable to attack.

2) It dangerously stresses both aircraft.

3) Being low exposes you to ground fire (Probably by from both sides IRL)

4) Being low cedes a tactical advantage to other enemies in the area.

5) The formation is weakened by your selfishness

It's perfectly reasonable to disregard these things in a simulation for the sake of fun, especially if you're an aggressive flyer who likes to notch up victories (Hellshade!). However, in doing so you're deliberately subjecting yourself to a quirk in the game that you probably shouldn't be flirting with too often for the sake of your virtual life. I rationalize this maneuver by the AI to be a desperate attempt to survive, preferring to take their chances with the factors listed above rather than continue to fight under unsatisfactory tactical conditions and, as such, I personally prefer to view my own excursions down after them as a tactical mistake when they do occur (Yes, I go through dry spells and get agitated enough to follow them down too.) Note that these lapses often result in the untimely death of my pilot by archie or by having to fight outnumbered in a damaged machine.

However, it is also possible that you may be in a formation whose leader initiates such a maneuver as a means of opening an attack. In which case it is understandable to either follow him or to descend at a more reasonable speed to provide a kind of "overwatch" for the members of the formation who do the actual engaging. (Not as much fun as shooting at them yourself but an important role nonetheless.)

I've also noticed that these dives often tend to be performed in a very and tight rapid spiral that, in it's own way, simulates an enemy feigning a spin to descend to low altitude. As such, I'd prefer that they remain since they are a part of what makes the combat in WoFF less of a tedious meat grinder, like some other simulations, and more of an authentic battle for survival. Generally I don't consider myself to be an apologist for things that are clearly unrealistic but, in this case, I choose the AI meteor maneuver over the AI meat grinder that would certainly ensue, were the AI to be deprived of this means of escape.

With that said, I think that the ideal compromise wouldn't be a mod that removes the power dive entirely or one that subjects the AI tot he same physics as the player at the expense of performance but, rather, an AI who is less likely to initiate such a maneuver.

BluckeyeBob,

I wasn't aware the AI and FMs had been encrypted. Unless modders can get access to them, I too see no way to change the AI Meteor Dive behavior. My guess (which is probably wrong) is that each of the various moves does indeed have a name and short "script" to follow where the AI decides based upon numerous variables and when things looks bad and they need to "get the hell out of Dodge City", they choose the Meteor Dive move. If we had access, then someone might be able to modify that move to somehow be less extreme. But as you say, they are encrypted. Maybe someday!

I do not think the AI FM is encrypted. It is just integral part of CFS3 code which we do not have access to. I assume the AI FM is quite different than f¨FM of human player, missing overspeed and overG control routines, so, for the sake of code simplicity, AIs do not need to be aware of these limits.
Just my 2 cents.

I know Rex coded the AI decisions so that rather than randomly choosing what they do, its based upon weighing a bunch of factors. I presume that means they have a series of "moves" to choose from. If there was a way to access the AI code, you might be able to simply block out the "Meteor Dive" move and redirect it to the fake death spiral.

Hellshade, that IS an intriguing possibility. Earlier, I misstated about the AI routines. What I meant is we don't have total access to all the AI decision-making and FM algorithms. However, I have glanced (but not recently) at the AI files that aren't encrypted. It's possible we might be able to mod them--how much, I don't know.

As a temporary, although simplistic and somewhat clunky workaround, you might experiment with turning on "AI fights to the death" in Workshop to see what effect it may have on the death-dive behavior. Unfortunately, turning on this option creates other problems, as I'm sure you are aware. But your above suggestion is interesting.

Well I'd probably bump up their morale modifier overall to be honest, generally they tend to run home to momma a little too frequently. (And I'm talking formations of scouts who panic and flee because they saw a 2 seater, it happens)

I will give that a try. If anything, it might tell us if that is the move the AI goes to when Morale fails completely. If it still happens, then maybe the move isn't tied to morale.

It would be great if you guys could take a look at the files. Obviously this behavior has been there since the beginning and isn't a "game breaker." It would just be cool to see it modded to somehow more realistically take into account the limits of the aircraft, if possible.

[Edit] Ace - bumping up their morale might delay the behavior long enough to get a longer lasting fight out of them for sure. Naturally, that increases the deadliness of the sim, given that they would fight longer. Even though I've only had one serious campaign pilot get near the 200 hour mark (most die between 40 and 100 hours), I would probably be okay with that. Maybe them fighting longer and harder would get ME to run away more often as my situation became more desperate!

The relevant file appears to be the simulation.xml file located in the main OBDWWI Over Flanders Fields folder. It contains information related to AI flight behavior while in formation and solo, as well as a list of various AI flight manuevers, such as pursuit and Chandelle, although there is only very limited information about how the AI might choose one manuever over another. Detailed information about that is likely buried in an encrypted file. One manuever in particular seems interesting to me. The manuever in question is "SpiralInF," although I have no idea what the "F" might mean. Fake spiral in, perhaps?

The formation flying section is interesting, although again it mainly only lists a few broad courses of action for the AI and the various factors (moral, damage, etc.) that influence these choices. For example, it appears (depending if the AI is flying a scout or a two-seater) that the AI has four broad choices when engaged with an enemy. They appear to be: "continue on," "withdraw," "run away," and "go home." The actual routines for these tactics appear to be buried in an encrypted file somewhere, so it's hard for me to say how "withdraw" differs from "run away" for example.

Maybe one of the devs would be so kind as to drop just a couple of very broad hints?

Great find, BB. I agree SpirallnF may well be the Fake Spiral (which I think is a great move the devs put in). I wonder what the difference between "withdraw" and "run away" are. "Run away" sounds more desperate, but that doesn't mean anything.

Do any of the manuevers sound like they could be the "Meteor Dive" move? I can't wait to get home and start looking. Awesome stuff.

The simulation.xml file has a setting for morale weight factor, which obviously controls how much influence a pilot's morale may have on the decision to fight or run away. Does anyone know the file that contains the actual morale for different classes of pilots? I bet it depends on a number of complex factors, such as experience and squadron quality, to name a few.

My guess is that withdraw refers to a tactical withdrawal. If the AI can regain an advantage, it may re-engage. If it's triggered to runaway, my guess is it will do it's best to "get the hell out of Dodge" and fly to the nearest airbase. This may even include starting with a desperate death-dive, but I can't know for sure without access to the actual routines.

I wonder if the sim has a debugging mode that might enable us to see some of these AI decisions in action? Hmm....

I *thought* that I saw in the Key Commands there was a version of the Lables that let you know what the AI was "thinking" at the time. I could be wrong.

The only manuever that suggests a death-dive while still fighting may be "reverseUnder" but I think it merely refers to a split-s manuever, rather than an actual dive. My suspicion is that the death-dive is one of the first moves that occurs when the AI has "decided" to runaway or go home, as I suggested in my previous post.

That makes sense. Bumping pilot Morale somehow may end up being the indirect way of pushing off the behavior itself if there's not a way to target that specific move. Of course, thats just another nut to crack but it may not be as challenging, hopefully.

There is a "labels activity" button, but the only activities I've seen are "taking-off," "in-transit," "fighting," "going home," and "landing." I don't recall ever seeing any other labels, but I may have missed them, particularly if I am in the middle of a dogfight!

I unmapped the Lables key altogether about 2 months ago so I couldn't "cheat" and see aircraft that were really far away or get any additional information about them. I do get jumped more often now, since I can't just speed over the the bad guys that the lables told me were there. LOL

BTW, that is why I have created mod to customize TAC and label colors. I set the opacity (alpha channel) of enemy labels to 1 (value of 0 is unacceptable by CFS3 engine) and enemy labels are invisible.
This mod is part of WOFF UE Multimod & Customiser.

You have made so many mods, sir. Can' tell you how much I appreciate your contributions. My favorite is still the Roll FM mod. The AI just look so much more natural flying with that.

I've done some testing with activity labels on and can confirm that the death-dive behavior occurs just after the AI takes more than insignificant damage. Although the activity labels still says "fighting" it always changes to "going home" immediately after the AI pulls out of the dive (there may also be a time delay before the label changes). The AI will then refuse to re-engage and will fly in a straight line to the nearest base unless it again begins to take damage from an enemy plane.

I seemed to have some minor success changing two variables in the simulation.xml file located in the main OBDWWI Over Flanders Fields folder. Scroll down to the section labeled "individual_fighter_v_fighter" and look for these variables:

and change to

This seems to make the AI stick around a little longer before bugging out and initiating the death-dive boogie. You can try other negative values, but always make sure that withdrawThreshold is only one value less than fightThreshold.

Careful! These results are extremely preliminary and require substantial additional testing! As always, be sure to back up the simulation.xml first before making any changes.

We may also be able to delay--but not eliminate--this problem by modifying the damage values in each aircraft's .xdp folder to make them more resistant to minor damage, although this would be a major project that would take a lot of time.

I will just look at playing with Threshold values first. Oddly enough though, I don't see a simulation.xml file in the CampaignData folder.

Found it. Excellent. Will give this a go later today!

I'm going to go ahead and hazard a guess here...

The meteor dive is not a maneuver at all, it doesn't correspond to anything in the simulation.xml file. My guess is that it is a result of the WAYPOINT that is set for the AI when the morale modifier reaches a set minimum.

Is there a way to test that, Ace? Its a good possibility.

Probably by checking "AI never backs down" in workshop.

I guess I meant is there way to make a mod that raises altitude of the waypoint for retreating aircraft. I know JJJ has a mod for setting waypoints and such, but I don't *think* it mods that waypoint.

Since it's a dynamically changing waypoint I'd say that it is attached to a trigger so probably not directly accessible but JJJ would be the man to ask.

I think it's probably neither. More likely, it is a hard coded feature left over from cfs3. All versions of OFF have had this behavior.

As a side note, this is why I was disappointed that OBD decided to lock some of the FM into encrypted files. With previous versions of OFF, you could modify settings such as parasitic drag, which if set correctly, slows aircraft dive speeds to more historical levels (remember the Tripe that could dive at speeds close to 300 mph)? Now, dive speed is limited by overspeed damage to the wing, but that unfortunately only seems to affect the human pilot.

There's no such thing as "Hard Coded". It's Software

Carefully reading your comments, guys, let me express my thoughts. The best and the most knowleable person to ask is WindMan. He knows the CFS3 code and he is the man involved in WOFF FM engine coding and implementation.
IMHO, the FM variables missing in individual *.xdp files (that were previously present in *.air files) were averaged for WWI planeset and hard coded (Softwared per AcePilto ) in CFS3 source code. Remaining variables are still present and accessible in aircraft respective *.xdp files. This solution resulted in simplified FM management, but, as adverse side effect, it prevents implementation oft some more complicated/different FMs, for example flyable Zeppelins or HandleyPage HP O/400 (as was promised in the past).
Just my 2 cents.

Quick confirmation that BuckeyeBob's tweaks suggested for the simulation.xml file indeed change AI behavior and seem to minimize death spirals. I've further tested the following values for the individual fighter vs. fighter setting in the main simulation.xml file (that resides in the OBDWW1 folder and controls quick combat missions):

fightThreshold="-3"
withdrawThreshold="-4"

Flying in an Alb. D.II against two Sopwith Pups, with AI skill set to "random," led to a prolonged fight down into the weeds, one Pup augering into the ground eventually while trying to loop too low, and the other barely ungluing itself from me before flying off. This leads me to assume that values of -4/-5 or lower would probably result in a fight to the death situation, even with sensible AI settings checked off in the workshop settings. Settings of -3/-4 still allow for some last-minute break in a dogfight, smaller numbers as BuckeyeBob suggested (-2/-3) give slightly less stubborn AI.

Happy flying,
Von S

No offence, guys, but you probaly missed the main point - can you chase the AI in dive without losing your wings?

Depends on how you view the AI's behavior. If you feel it is historical, than by all means we should be looking at ways to follow without losing your own wings. If not historical, we should be looking at ways to reduce or eliminate the AI's behavior. That said, if you can work one of your miracles and find a solution so the AI has to obey the laws of physics just like the human pilots, then that would be fantastic!

Was able to chase them down easily in the Alb. D.I and D.II - other crates may not fare as well. Although, I didn't notice any of those death-defying dives by the AI using the -3/-4 threshold settings I'm tinkering with currently, or with BuckeyeBob's -2/-3 settings. More thorough testing is of course needed. I look forward to further progress on this topic.

Von S

I can confirm that a few hours worth of flying in a DI today and none of the enemy AI made the meteor dive move with BuckeyeBods -2/-3 setting. As VonS says, more testing is needed and I am not sure what else might be being affected in a negative way, but it's a promising start.

BB, are you sure WOFF reads and uses data from this file (location)? I tried to delete it (for testing purposes) and WOFF have not complained and run campaign mission as usually.
I suppose we should edit/modify simulation.xml file located at "xx:\OBDSoftware\WOFF\OBDWW1 Over Flanders Fields" folder. File located at Default folder is old, obsolete version, incompatible with current version of WOFF UE.
BTW, this version allowed to fly FE2b in fighter role, enabling following attributes:

That option is unfortunately not available anymore and FE2bs can be used as moving targets only .

Yes, there are two simulation.xml files. One is located in the main Over Flanders Fields folder and is labeled for QC battles, and the other is in the campaigns\CampaignData\Default folder and is (presumably) for campaign missions. Whether the latter has been deprecated or not, I do not know. Perhaps if WOFF doesn't find the simulation file located in the \Default folder, it loads the simulation file from the main folder? To be on the safe side, perhaps we should alter the two values I was discussing in both files.

Edit: upon further investigation, it appears that Jara is right--the simulation file in the campaigns folder is ignored (sorry that I ever doubted you, Jara)! (Edit: I have changed my previous posts to reflect the correct information about which simulation.xml file to change).


What would happen if you copied these to the simulation file in the main folder--anything? Or is it ignored?

Sorry, I have not noticed your reply and question.
I tried that, of course, and game complained due to unknown attributes - this part of code has been removed by DEVs (probably because Gothas forgot about their bomber role and played fighters ?).

bump to thread...

A bit of shameless self-promotion here to remind folks that you can adjust the AI aggressiveness a bit by manipulating the simulation.xml file in the main WOFF folder, OR you can use my "Simple Little AI mod" available here to change the fightThreshold and withdrawThreshold values.

Be sure to read the brief readme file included with the mod and backup the simulation.xml file if you decide to try and modify the file yourself.